Arbitrary choices in theoretical reasoning?

Some writers (e.g. Harman) say that arbitrary choices are sometimes reasonable in practical reasoning, but never in theoretical reasoning. So when choosing which baked bean can to take off the shelf at the supermarket, it's ok to just randomly pick any old one. In contrast, there is no piece of reasoning where it's ok to just randomly pick something to believe.

You might think that it's false that arbitrary choices are rational in practical reasoning (perhaps you ought to take the closest and least-dented can). But even if that's true, here's an example that might be a case of rational and arbitrary theoretical reasoning.

I'm deciding whether or not to believe the proposition:
P: By the time I finish this piece of reasoning, I will believe P.

I take an arbitrary choice to be one where where doing nothing is worse than some options available, and where there is more than one jointly best option. This is true of this piece of reasoning. If you believe that P is true, then P is true, and so you have one more correct belief. If you believe that P is false, then P is false, and so you have one more correct belief. So these options are equally good. And each is better than not believing either P or its negation. To withhold belief here is just to miss out on a 'free' true belief.
(I mean that not believing this is irrational to some extent. Obviously people have better things to do with their time than believe such bizarre propositins, and so this extremely low level of irrationality is outweighed by the vastly more important things we can devote mental energy to.)